ROC Curve Analysis for Diagnostic Tests
In a research for investigating the performance of two diagnostic tests, reading scores were generated and placed in the following data file along with the patients’ disease status. The two test procedures are called Test 1 and Test 2 are proposed for detecting a certain disease, called disease A. The purpose of these two test s are to identify whether patient under the test having disease A or not. It is important to rule-in those patients who have the disease because it matters live-and-death for the patients. The data is stored at: http://people.ysu.edu/~gchang/stat/diagnostictests.sav .
Please run the basic analysis using the data and write a report to give the researchers some suggestions based on these data. In the report, you should include the result of your analysis by attaching important SPSS output to support your writing, recommend a better test among the two under the study for classifying patients based on observing the ROC curves, and also provide a criterion (cut-off point) for ruling-in those patients who are very likely to have the disease. You are not an expert in the medical field therefore suggesting the cut-off is definitely not what you can do especially without knowing some details of the disease. I would like you just select a cut-off and explain your own reason for choosing it, of course, also based on the sensitivity and specificity from the data. Please use the Internet to search for some articles that talk about the sensitivity analysis and the use of ROC to help you learn more about the method. The ROC function in SPSS is limited. I don’t think it has the test for comparing test procedures. It would be GREAT if you can find any free software online for testing statistical significant difference between the two tests! But, it is not required. If you find any articles that help you in the analysis in this report, please cite the reference in your writing and also list them at the end of your report.
A reference about sensitivity analysis and ROC: (http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2556590/pdf/cbr29_s_pgs83.pdf)